Differents results for margin and allmargin. close loop stability

Hello

I have modelinzing a complex model using simulink.

I linearize it using linearize commande (using the good linear analisys points, so I have a SISO system) , and then I am using margin or allmargin to obtain stability criterias.

and when I am using margin or allmargin I do not find the same results :

K>> [MG, PM] = margin(sys_linearize)

MG =

3.4953

PM =

87.6479

K>> s=allmargin(sys_linearize)

s =

struct with fields:

GainMargin: [0.1160 3.4953 7.2396 10.8338 14.6787 18.5872 22.4725 26.4404 31.1912 40.5220 275.6987 116.8241]

GMFrequency: [0.0208 1.4592 4.6862 7.7867 10.7234 13.4281 15.8367 17.8916 19.5435 20.7531 21.4934 104.7198]

PhaseMargin: 87.6479

PMFrequency: 0.2367

DelayMargin: 215.4141

DMFrequency: 0.2367

Stable: 0

you can see figures also, made with bodbplot and using menu "caracteristics\minimum stability margin" and "caracteristics\all stability margin" which isllustrate my question (see attachment)

• so margin say GainMargin = 20*log10(3.4953) = 10.9 dB, which should mean that close loop is stable (PhaseMargin = 87°)
• allmargin say GainMargin at = 20*log10(0.1160) = -18.7 dB lowest frequency (so the first time phase is crossing -180°), which should mean that close loop is unstable
• on bodeplot figure and "caracteristics\minimum stability margin" diplay, we have MG = 10.9 dB and PM = 87° and close loop unstable !

I do not undertsand those differences

I have a simulation using simulink for this operating point showing self-sustaining oscillation ... so I think that my system with close loop is unstable for this operating point and so I should retain GainMargin = -18.7 dB and PhaseMargin = 87°

if needed, you can find with attachements a mat file containing le LT system

thank you in advance

Fabien